Reduced Order Non-INtrusive (RONIN) Modeling for Strategic Defense Planning
Author(s)
Bateman, Mark
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
The Department of Defense (DoD) along with research organizations like RAND have documented a strategic gap in the ability to conduct exploratory analyses to support capability development that seek to exploit both technological and doctrinal conceptual solutions. There are many facets related to conducting exploratory analyses, from examining trends in technology development to making intelligence assessments of potential adversary capabilities; in the end, efforts to create models and simulations to explore various scenarios lie at the heart of providing senior leaders analytical support. In this work, the overarching strategic gap was decomposed into more focused areas of needed research, starting with an exploration of current methods for integrating different models in order to meet the concerns of Congress with regards to quality, accuracy, and dependability; noting that they have become too computationally prohibitive for exploring a large design or decision space. Further observations noted that the expected value of performance metrics requiring complex and potentially nonlinear models to quantify does not provide sufficient traceability when used between different levels of model abstraction. Additionally, current model abstraction methods have difficulty accounting for the increasing dimensionality associated with increasingly complex models or simulations. These observations lead to the objective of this research, which is the formulation and demonstration of a methodology which leverages reduced order modeling (ROM) methods for traceable model abstraction that effectively and efficiently captures complex system-of-systems behaviors within current military operations modeling and simulation methods.
A review of current literature led to the derivation of the following requirements for ROMs: there is a need to account for nonlinear interdependencies, underlying physical phenomena, and stochastic effects. A set of research questions, hypotheses, and experiments were posed and completed to further understand and address identified gaps. All of which guided the formulation of the Reduced Order Non-INtrusive (RONIN) modeling methodology, enabling the accomplishment of the stated research objective. The RONIN modeling methodology creates and implements predictive reduced order surrogate models capturing more information regarding behaviors and interactions as compared to traditional methods such as “look-up” tables or simple passing of expected values. Finally, to demonstrate the RONIN modeling methodology’s ability the meet the research objective, a notional United States Air Force use case was defined, and a DoD standard simulation framework was used to generate a Full Order Model (FOM) which output a set of response distributions. Responses from mission-level models work to quantify system-of-systems behavior and can range from simple counts of assets, weapons, or fuel, to advanced metrics which aim to calculate operational effectiveness. This use case modeled a Suppression of Enemy Air Defense (SEAD) mission exploring how different decisions and force structures affect the total number of friendly losses and enemy kills. Ultimately, the RONIN modeling method was used to create a predictive surrogate model which was able to reconstruct output distributions which are statically consistent with the original FOM output data.
Sponsor
Date
2023-07-10
Extent
Resource Type
Text
Resource Subtype
Dissertation